2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <kandros@umich.edu>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/file.h>
45 #include <linux/mount.h>
46 #include <linux/workqueue.h>
47 #include <linux/smp_lock.h>
48 #include <linux/kthread.h>
49 #include <linux/nfs4.h>
50 #include <linux/nfsd/state.h>
51 #include <linux/nfsd/xdr4.h>
52 #include <linux/namei.h>
53 #include <linux/swap.h>
54 #include <linux/mutex.h>
55 #include <linux/lockd/bind.h>
56 #include <linux/module.h>
57 #include <linux/sunrpc/svcauth_gss.h>
59 #define NFSDDBG_FACILITY NFSDDBG_PROC
62 static time_t lease_time
= 90; /* default lease time */
63 static time_t user_lease_time
= 90;
64 static time_t boot_time
;
65 static u32 current_ownerid
= 1;
66 static u32 current_fileid
= 1;
67 static u32 current_delegid
= 1;
69 static stateid_t zerostateid
; /* bits all 0 */
70 static stateid_t onestateid
; /* bits all 1 */
71 static u64 current_sessionid
= 1;
73 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
74 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
76 /* forward declarations */
77 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
78 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
79 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
80 static void nfs4_set_recdir(char *recdir
);
84 /* Currently used for almost all code touching nfsv4 state: */
85 static DEFINE_MUTEX(client_mutex
);
88 * Currently used for the del_recall_lru and file hash table. In an
89 * effort to decrease the scope of the client_mutex, this spinlock may
90 * eventually cover more:
92 static DEFINE_SPINLOCK(recall_lock
);
94 static struct kmem_cache
*stateowner_slab
= NULL
;
95 static struct kmem_cache
*file_slab
= NULL
;
96 static struct kmem_cache
*stateid_slab
= NULL
;
97 static struct kmem_cache
*deleg_slab
= NULL
;
100 nfs4_lock_state(void)
102 mutex_lock(&client_mutex
);
106 nfs4_unlock_state(void)
108 mutex_unlock(&client_mutex
);
112 opaque_hashval(const void *ptr
, int nbytes
)
114 unsigned char *cptr
= (unsigned char *) ptr
;
124 static struct list_head del_recall_lru
;
127 put_nfs4_file(struct nfs4_file
*fi
)
129 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
130 list_del(&fi
->fi_hash
);
131 spin_unlock(&recall_lock
);
133 kmem_cache_free(file_slab
, fi
);
138 get_nfs4_file(struct nfs4_file
*fi
)
140 atomic_inc(&fi
->fi_ref
);
143 static int num_delegations
;
144 unsigned int max_delegations
;
147 * Open owner state (share locks)
150 /* hash tables for nfs4_stateowner */
151 #define OWNER_HASH_BITS 8
152 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
153 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
155 #define ownerid_hashval(id) \
156 ((id) & OWNER_HASH_MASK)
157 #define ownerstr_hashval(clientid, ownername) \
158 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
160 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
161 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
163 /* hash table for nfs4_file */
164 #define FILE_HASH_BITS 8
165 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
166 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
167 /* hash table for (open)nfs4_stateid */
168 #define STATEID_HASH_BITS 10
169 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
170 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
172 #define file_hashval(x) \
173 hash_ptr(x, FILE_HASH_BITS)
174 #define stateid_hashval(owner_id, file_id) \
175 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
177 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
178 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
180 static struct nfs4_delegation
*
181 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
183 struct nfs4_delegation
*dp
;
184 struct nfs4_file
*fp
= stp
->st_file
;
185 struct nfs4_cb_conn
*cb
= &stp
->st_stateowner
->so_client
->cl_cb_conn
;
187 dprintk("NFSD alloc_init_deleg\n");
188 if (fp
->fi_had_conflict
)
190 if (num_delegations
> max_delegations
)
192 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
196 INIT_LIST_HEAD(&dp
->dl_perfile
);
197 INIT_LIST_HEAD(&dp
->dl_perclnt
);
198 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
203 get_file(stp
->st_vfs_file
);
204 dp
->dl_vfs_file
= stp
->st_vfs_file
;
206 dp
->dl_ident
= cb
->cb_ident
;
207 dp
->dl_stateid
.si_boot
= get_seconds();
208 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
209 dp
->dl_stateid
.si_fileid
= 0;
210 dp
->dl_stateid
.si_generation
= 0;
211 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
213 atomic_set(&dp
->dl_count
, 1);
214 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
215 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
220 nfs4_put_delegation(struct nfs4_delegation
*dp
)
222 if (atomic_dec_and_test(&dp
->dl_count
)) {
223 dprintk("NFSD: freeing dp %p\n",dp
);
224 put_nfs4_file(dp
->dl_file
);
225 kmem_cache_free(deleg_slab
, dp
);
230 /* Remove the associated file_lock first, then remove the delegation.
231 * lease_modify() is called to remove the FS_LEASE file_lock from
232 * the i_flock list, eventually calling nfsd's lock_manager
233 * fl_release_callback.
236 nfs4_close_delegation(struct nfs4_delegation
*dp
)
238 struct file
*filp
= dp
->dl_vfs_file
;
240 dprintk("NFSD: close_delegation dp %p\n",dp
);
241 dp
->dl_vfs_file
= NULL
;
242 /* The following nfsd_close may not actually close the file,
243 * but we want to remove the lease in any case. */
245 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
249 /* Called under the state lock. */
251 unhash_delegation(struct nfs4_delegation
*dp
)
253 list_del_init(&dp
->dl_perfile
);
254 list_del_init(&dp
->dl_perclnt
);
255 spin_lock(&recall_lock
);
256 list_del_init(&dp
->dl_recall_lru
);
257 spin_unlock(&recall_lock
);
258 nfs4_close_delegation(dp
);
259 nfs4_put_delegation(dp
);
266 /* Hash tables for nfs4_clientid state */
267 #define CLIENT_HASH_BITS 4
268 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
269 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
271 #define clientid_hashval(id) \
272 ((id) & CLIENT_HASH_MASK)
273 #define clientstr_hashval(name) \
274 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
276 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
277 * used in reboot/reset lease grace period processing
279 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
280 * setclientid_confirmed info.
282 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
285 * client_lru holds client queue ordered by nfs4_client.cl_time
288 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
289 * for last close replay.
291 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
292 static int reclaim_str_hashtbl_size
= 0;
293 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
294 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
295 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
296 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
297 static struct list_head client_lru
;
298 static struct list_head close_lru
;
300 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
302 list_del(&stp
->st_hash
);
303 list_del(&stp
->st_perfile
);
304 list_del(&stp
->st_perstateowner
);
307 static void free_generic_stateid(struct nfs4_stateid
*stp
)
309 put_nfs4_file(stp
->st_file
);
310 kmem_cache_free(stateid_slab
, stp
);
313 static void release_lock_stateid(struct nfs4_stateid
*stp
)
315 unhash_generic_stateid(stp
);
316 locks_remove_posix(stp
->st_vfs_file
, (fl_owner_t
)stp
->st_stateowner
);
317 free_generic_stateid(stp
);
320 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
322 struct nfs4_stateid
*stp
;
324 list_del(&sop
->so_idhash
);
325 list_del(&sop
->so_strhash
);
326 list_del(&sop
->so_perstateid
);
327 while (!list_empty(&sop
->so_stateids
)) {
328 stp
= list_first_entry(&sop
->so_stateids
,
329 struct nfs4_stateid
, st_perstateowner
);
330 release_lock_stateid(stp
);
334 static void release_lockowner(struct nfs4_stateowner
*sop
)
336 unhash_lockowner(sop
);
337 nfs4_put_stateowner(sop
);
341 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
343 struct nfs4_stateowner
*lock_sop
;
345 while (!list_empty(&open_stp
->st_lockowners
)) {
346 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
347 struct nfs4_stateowner
, so_perstateid
);
348 /* list_del(&open_stp->st_lockowners); */
349 BUG_ON(lock_sop
->so_is_open_owner
);
350 release_lockowner(lock_sop
);
354 static void release_open_stateid(struct nfs4_stateid
*stp
)
356 unhash_generic_stateid(stp
);
357 release_stateid_lockowners(stp
);
358 nfsd_close(stp
->st_vfs_file
);
359 free_generic_stateid(stp
);
362 static void unhash_openowner(struct nfs4_stateowner
*sop
)
364 struct nfs4_stateid
*stp
;
366 list_del(&sop
->so_idhash
);
367 list_del(&sop
->so_strhash
);
368 list_del(&sop
->so_perclient
);
369 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
370 while (!list_empty(&sop
->so_stateids
)) {
371 stp
= list_first_entry(&sop
->so_stateids
,
372 struct nfs4_stateid
, st_perstateowner
);
373 release_open_stateid(stp
);
377 static void release_openowner(struct nfs4_stateowner
*sop
)
379 unhash_openowner(sop
);
380 list_del(&sop
->so_close_lru
);
381 nfs4_put_stateowner(sop
);
384 static DEFINE_SPINLOCK(sessionid_lock
);
385 #define SESSION_HASH_SIZE 512
386 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
389 hash_sessionid(struct nfs4_sessionid
*sessionid
)
391 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
393 return sid
->sequence
% SESSION_HASH_SIZE
;
397 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
399 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
400 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
404 gen_sessionid(struct nfsd4_session
*ses
)
406 struct nfs4_client
*clp
= ses
->se_client
;
407 struct nfsd4_sessionid
*sid
;
409 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
410 sid
->clientid
= clp
->cl_clientid
;
411 sid
->sequence
= current_sessionid
++;
416 * Give the client the number of slots it requests bound by
417 * NFSD_MAX_SLOTS_PER_SESSION and by sv_drc_max_pages.
419 * If we run out of pages (sv_drc_pages_used == sv_drc_max_pages) we
420 * should (up to a point) re-negotiate active sessions and reduce their
421 * slot usage to make rooom for new connections. For now we just fail the
424 static int set_forechannel_maxreqs(struct nfsd4_channel_attrs
*fchan
)
426 int status
= 0, np
= fchan
->maxreqs
* NFSD_PAGES_PER_SLOT
;
428 spin_lock(&nfsd_serv
->sv_lock
);
429 if (np
+ nfsd_serv
->sv_drc_pages_used
> nfsd_serv
->sv_drc_max_pages
)
430 np
= nfsd_serv
->sv_drc_max_pages
- nfsd_serv
->sv_drc_pages_used
;
431 nfsd_serv
->sv_drc_pages_used
+= np
;
432 spin_unlock(&nfsd_serv
->sv_lock
);
435 status
= nfserr_resource
;
438 fchan
->maxreqs
= np
/ NFSD_PAGES_PER_SLOT
;
444 * fchan holds the client values on input, and the server values on output
446 static int init_forechannel_attrs(struct svc_rqst
*rqstp
,
447 struct nfsd4_session
*session
,
448 struct nfsd4_channel_attrs
*fchan
)
451 __u32 maxcount
= svc_max_payload(rqstp
);
453 /* headerpadsz set to zero in encode routine */
455 /* Use the client's max request and max response size if possible */
456 if (fchan
->maxreq_sz
> maxcount
)
457 fchan
->maxreq_sz
= maxcount
;
458 session
->se_fmaxreq_sz
= fchan
->maxreq_sz
;
460 if (fchan
->maxresp_sz
> maxcount
)
461 fchan
->maxresp_sz
= maxcount
;
462 session
->se_fmaxresp_sz
= fchan
->maxresp_sz
;
464 /* Set the max response cached size our default which is
465 * a multiple of PAGE_SIZE and small */
466 session
->se_fmaxresp_cached
= NFSD_PAGES_PER_SLOT
* PAGE_SIZE
;
467 fchan
->maxresp_cached
= session
->se_fmaxresp_cached
;
469 /* Use the client's maxops if possible */
470 if (fchan
->maxops
> NFSD_MAX_OPS_PER_COMPOUND
)
471 fchan
->maxops
= NFSD_MAX_OPS_PER_COMPOUND
;
472 session
->se_fmaxops
= fchan
->maxops
;
474 /* try to use the client requested number of slots */
475 if (fchan
->maxreqs
> NFSD_MAX_SLOTS_PER_SESSION
)
476 fchan
->maxreqs
= NFSD_MAX_SLOTS_PER_SESSION
;
478 /* FIXME: Error means no more DRC pages so the server should
479 * recover pages from existing sessions. For now fail session
482 status
= set_forechannel_maxreqs(fchan
);
484 session
->se_fnumslots
= fchan
->maxreqs
;
489 alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
,
490 struct nfsd4_create_session
*cses
)
492 struct nfsd4_session
*new, tmp
;
493 int idx
, status
= nfserr_resource
, slotsize
;
495 memset(&tmp
, 0, sizeof(tmp
));
497 /* FIXME: For now, we just accept the client back channel attributes. */
498 status
= init_forechannel_attrs(rqstp
, &tmp
, &cses
->fore_channel
);
502 /* allocate struct nfsd4_session and slot table in one piece */
503 slotsize
= tmp
.se_fnumslots
* sizeof(struct nfsd4_slot
);
504 new = kzalloc(sizeof(*new) + slotsize
, GFP_KERNEL
);
508 memcpy(new, &tmp
, sizeof(*new));
510 new->se_client
= clp
;
512 idx
= hash_sessionid(&new->se_sessionid
);
513 memcpy(clp
->cl_sessionid
.data
, new->se_sessionid
.data
,
514 NFS4_MAX_SESSIONID_LEN
);
516 new->se_flags
= cses
->flags
;
517 kref_init(&new->se_ref
);
518 spin_lock(&sessionid_lock
);
519 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
520 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
521 spin_unlock(&sessionid_lock
);
528 /* caller must hold sessionid_lock */
529 static struct nfsd4_session
*
530 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
532 struct nfsd4_session
*elem
;
535 dump_sessionid(__func__
, sessionid
);
536 idx
= hash_sessionid(sessionid
);
537 dprintk("%s: idx is %d\n", __func__
, idx
);
538 /* Search in the appropriate list */
539 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
540 dump_sessionid("list traversal", &elem
->se_sessionid
);
541 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
542 NFS4_MAX_SESSIONID_LEN
)) {
547 dprintk("%s: session not found\n", __func__
);
551 /* caller must hold sessionid_lock */
553 unhash_session(struct nfsd4_session
*ses
)
555 list_del(&ses
->se_hash
);
556 list_del(&ses
->se_perclnt
);
560 release_session(struct nfsd4_session
*ses
)
562 spin_lock(&sessionid_lock
);
564 spin_unlock(&sessionid_lock
);
565 nfsd4_put_session(ses
);
568 static void nfsd4_release_respages(struct page
**respages
, short resused
);
571 free_session(struct kref
*kref
)
573 struct nfsd4_session
*ses
;
576 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
577 for (i
= 0; i
< ses
->se_fnumslots
; i
++) {
578 struct nfsd4_cache_entry
*e
= &ses
->se_slots
[i
].sl_cache_entry
;
579 nfsd4_release_respages(e
->ce_respages
, e
->ce_resused
);
585 renew_client(struct nfs4_client
*clp
)
588 * Move client to the end to the LRU list.
590 dprintk("renewing client (clientid %08x/%08x)\n",
591 clp
->cl_clientid
.cl_boot
,
592 clp
->cl_clientid
.cl_id
);
593 list_move_tail(&clp
->cl_lru
, &client_lru
);
594 clp
->cl_time
= get_seconds();
597 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
599 STALE_CLIENTID(clientid_t
*clid
)
601 if (clid
->cl_boot
== boot_time
)
603 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
604 clid
->cl_boot
, clid
->cl_id
, boot_time
);
609 * XXX Should we use a slab cache ?
610 * This type of memory management is somewhat inefficient, but we use it
611 * anyway since SETCLIENTID is not a common operation.
613 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
615 struct nfs4_client
*clp
;
617 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
620 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
621 if (clp
->cl_name
.data
== NULL
) {
625 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
626 clp
->cl_name
.len
= name
.len
;
631 shutdown_callback_client(struct nfs4_client
*clp
)
633 struct rpc_clnt
*clnt
= clp
->cl_cb_conn
.cb_client
;
637 * Callback threads take a reference on the client, so there
638 * should be no outstanding callbacks at this point.
640 clp
->cl_cb_conn
.cb_client
= NULL
;
641 rpc_shutdown_client(clnt
);
643 if (clp
->cl_cb_conn
.cb_cred
) {
644 put_rpccred(clp
->cl_cb_conn
.cb_cred
);
645 clp
->cl_cb_conn
.cb_cred
= NULL
;
650 free_client(struct nfs4_client
*clp
)
652 shutdown_callback_client(clp
);
653 nfsd4_release_respages(clp
->cl_slot
.sl_cache_entry
.ce_respages
,
654 clp
->cl_slot
.sl_cache_entry
.ce_resused
);
655 if (clp
->cl_cred
.cr_group_info
)
656 put_group_info(clp
->cl_cred
.cr_group_info
);
657 kfree(clp
->cl_principal
);
658 kfree(clp
->cl_name
.data
);
663 put_nfs4_client(struct nfs4_client
*clp
)
665 if (atomic_dec_and_test(&clp
->cl_count
))
670 expire_client(struct nfs4_client
*clp
)
672 struct nfs4_stateowner
*sop
;
673 struct nfs4_delegation
*dp
;
674 struct list_head reaplist
;
676 dprintk("NFSD: expire_client cl_count %d\n",
677 atomic_read(&clp
->cl_count
));
679 INIT_LIST_HEAD(&reaplist
);
680 spin_lock(&recall_lock
);
681 while (!list_empty(&clp
->cl_delegations
)) {
682 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
683 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
685 list_del_init(&dp
->dl_perclnt
);
686 list_move(&dp
->dl_recall_lru
, &reaplist
);
688 spin_unlock(&recall_lock
);
689 while (!list_empty(&reaplist
)) {
690 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
691 list_del_init(&dp
->dl_recall_lru
);
692 unhash_delegation(dp
);
694 list_del(&clp
->cl_idhash
);
695 list_del(&clp
->cl_strhash
);
696 list_del(&clp
->cl_lru
);
697 while (!list_empty(&clp
->cl_openowners
)) {
698 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
699 release_openowner(sop
);
701 while (!list_empty(&clp
->cl_sessions
)) {
702 struct nfsd4_session
*ses
;
703 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
705 release_session(ses
);
707 put_nfs4_client(clp
);
710 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
)
712 struct nfs4_client
*clp
;
714 clp
= alloc_client(name
);
717 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
718 atomic_set(&clp
->cl_count
, 1);
719 atomic_set(&clp
->cl_cb_conn
.cb_set
, 0);
720 INIT_LIST_HEAD(&clp
->cl_idhash
);
721 INIT_LIST_HEAD(&clp
->cl_strhash
);
722 INIT_LIST_HEAD(&clp
->cl_openowners
);
723 INIT_LIST_HEAD(&clp
->cl_delegations
);
724 INIT_LIST_HEAD(&clp
->cl_sessions
);
725 INIT_LIST_HEAD(&clp
->cl_lru
);
729 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
731 memcpy(target
->cl_verifier
.data
, source
->data
,
732 sizeof(target
->cl_verifier
.data
));
735 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
737 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
738 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
741 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
743 target
->cr_uid
= source
->cr_uid
;
744 target
->cr_gid
= source
->cr_gid
;
745 target
->cr_group_info
= source
->cr_group_info
;
746 get_group_info(target
->cr_group_info
);
749 static int same_name(const char *n1
, const char *n2
)
751 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
755 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
757 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
761 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
763 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
766 /* XXX what about NGROUP */
768 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
770 return cr1
->cr_uid
== cr2
->cr_uid
;
773 static void gen_clid(struct nfs4_client
*clp
)
775 static u32 current_clientid
= 1;
777 clp
->cl_clientid
.cl_boot
= boot_time
;
778 clp
->cl_clientid
.cl_id
= current_clientid
++;
781 static void gen_confirm(struct nfs4_client
*clp
)
786 p
= (u32
*)clp
->cl_confirm
.data
;
787 *p
++ = get_seconds();
791 static int check_name(struct xdr_netobj name
)
795 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
796 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
803 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
805 unsigned int idhashval
;
807 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
808 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
809 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
810 list_add_tail(&clp
->cl_lru
, &client_lru
);
811 clp
->cl_time
= get_seconds();
815 move_to_confirmed(struct nfs4_client
*clp
)
817 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
818 unsigned int strhashval
;
820 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
821 list_del_init(&clp
->cl_strhash
);
822 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
823 strhashval
= clientstr_hashval(clp
->cl_recdir
);
824 list_add(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
828 static struct nfs4_client
*
829 find_confirmed_client(clientid_t
*clid
)
831 struct nfs4_client
*clp
;
832 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
834 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
835 if (same_clid(&clp
->cl_clientid
, clid
))
841 static struct nfs4_client
*
842 find_unconfirmed_client(clientid_t
*clid
)
844 struct nfs4_client
*clp
;
845 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
847 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
848 if (same_clid(&clp
->cl_clientid
, clid
))
855 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
856 * parameter. Matching is based on the fact the at least one of the
857 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
859 * FIXME: we need to unify the clientid namespaces for nfsv4.x
860 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
861 * and SET_CLIENTID{,_CONFIRM}
864 match_clientid_establishment(struct nfs4_client
*clp
, bool use_exchange_id
)
866 bool has_exchange_flags
= (clp
->cl_exchange_flags
!= 0);
867 return use_exchange_id
== has_exchange_flags
;
870 static struct nfs4_client
*
871 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
,
872 bool use_exchange_id
)
874 struct nfs4_client
*clp
;
876 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
877 if (same_name(clp
->cl_recdir
, dname
) &&
878 match_clientid_establishment(clp
, use_exchange_id
))
884 static struct nfs4_client
*
885 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
,
886 bool use_exchange_id
)
888 struct nfs4_client
*clp
;
890 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
891 if (same_name(clp
->cl_recdir
, dname
) &&
892 match_clientid_establishment(clp
, use_exchange_id
))
898 /* a helper function for parse_callback */
900 parse_octet(unsigned int *lenp
, char **addrp
)
902 unsigned int len
= *lenp
;
914 if ((c
< '0') || (c
> '9')) {
920 n
= (n
* 10) + (c
- '0');
931 /* parse and set the setclientid ipv4 callback address */
933 parse_ipv4(unsigned int addr_len
, char *addr_val
, unsigned int *cbaddrp
, unsigned short *cbportp
)
938 u32 addrlen
= addr_len
;
939 char *addr
= addr_val
;
944 for(i
= 4; i
> 0 ; i
--) {
945 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
948 cbaddr
|= (temp
<< shift
);
956 for(i
= 2; i
> 0 ; i
--) {
957 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
960 cbport
|= (temp
<< shift
);
969 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
)
971 struct nfs4_cb_conn
*cb
= &clp
->cl_cb_conn
;
973 /* Currently, we only support tcp for the callback channel */
974 if ((se
->se_callback_netid_len
!= 3) || memcmp((char *)se
->se_callback_netid_val
, "tcp", 3))
977 if ( !(parse_ipv4(se
->se_callback_addr_len
, se
->se_callback_addr_val
,
978 &cb
->cb_addr
, &cb
->cb_port
)))
980 cb
->cb_prog
= se
->se_callback_prog
;
981 cb
->cb_ident
= se
->se_callback_ident
;
984 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
985 "will not receive delegations\n",
986 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
992 nfsd4_set_statp(struct svc_rqst
*rqstp
, __be32
*statp
)
994 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
996 resp
->cstate
.statp
= statp
;
1000 * Dereference the result pages.
1003 nfsd4_release_respages(struct page
**respages
, short resused
)
1007 dprintk("--> %s\n", __func__
);
1008 for (i
= 0; i
< resused
; i
++) {
1011 put_page(respages
[i
]);
1017 nfsd4_copy_pages(struct page
**topages
, struct page
**frompages
, short count
)
1021 for (i
= 0; i
< count
; i
++) {
1022 topages
[i
] = frompages
[i
];
1025 get_page(topages
[i
]);
1030 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1031 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1032 * length of the XDR response is less than se_fmaxresp_cached
1033 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1034 * of the reply (e.g. readdir).
1036 * Store the base and length of the rq_req.head[0] page
1037 * of the NFSv4.1 data, just past the rpc header.
1040 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1042 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
1043 struct svc_rqst
*rqstp
= resp
->rqstp
;
1044 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1045 struct nfsd4_op
*op
= &args
->ops
[resp
->opcnt
];
1046 struct kvec
*resv
= &rqstp
->rq_res
.head
[0];
1048 dprintk("--> %s entry %p\n", __func__
, entry
);
1050 /* Don't cache a failed OP_SEQUENCE. */
1051 if (resp
->opcnt
== 1 && op
->opnum
== OP_SEQUENCE
&& resp
->cstate
.status
)
1054 nfsd4_release_respages(entry
->ce_respages
, entry
->ce_resused
);
1055 entry
->ce_opcnt
= resp
->opcnt
;
1056 entry
->ce_status
= resp
->cstate
.status
;
1059 * Don't need a page to cache just the sequence operation - the slot
1063 if (nfsd4_not_cached(resp
)) {
1064 entry
->ce_resused
= 0;
1065 entry
->ce_rpchdrlen
= 0;
1066 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__
,
1067 resp
->cstate
.slot
->sl_cache_entry
.ce_cachethis
);
1070 entry
->ce_resused
= rqstp
->rq_resused
;
1071 if (entry
->ce_resused
> NFSD_PAGES_PER_SLOT
+ 1)
1072 entry
->ce_resused
= NFSD_PAGES_PER_SLOT
+ 1;
1073 nfsd4_copy_pages(entry
->ce_respages
, rqstp
->rq_respages
,
1075 entry
->ce_datav
.iov_base
= resp
->cstate
.statp
;
1076 entry
->ce_datav
.iov_len
= resv
->iov_len
- ((char *)resp
->cstate
.statp
-
1077 (char *)page_address(rqstp
->rq_respages
[0]));
1078 /* Current request rpc header length*/
1079 entry
->ce_rpchdrlen
= (char *)resp
->cstate
.statp
-
1080 (char *)page_address(rqstp
->rq_respages
[0]);
1084 * We keep the rpc header, but take the nfs reply from the replycache.
1087 nfsd41_copy_replay_data(struct nfsd4_compoundres
*resp
,
1088 struct nfsd4_cache_entry
*entry
)
1090 struct svc_rqst
*rqstp
= resp
->rqstp
;
1091 struct kvec
*resv
= &resp
->rqstp
->rq_res
.head
[0];
1094 /* Current request rpc header length*/
1095 len
= (char *)resp
->cstate
.statp
-
1096 (char *)page_address(rqstp
->rq_respages
[0]);
1097 if (entry
->ce_datav
.iov_len
+ len
> PAGE_SIZE
) {
1098 dprintk("%s v41 cached reply too large (%Zd).\n", __func__
,
1099 entry
->ce_datav
.iov_len
);
1102 /* copy the cached reply nfsd data past the current rpc header */
1103 memcpy((char *)resv
->iov_base
+ len
, entry
->ce_datav
.iov_base
,
1104 entry
->ce_datav
.iov_len
);
1105 resv
->iov_len
= len
+ entry
->ce_datav
.iov_len
;
1110 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1111 * cached page. Replace any futher replay pages from the cache.
1114 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1115 struct nfsd4_sequence
*seq
)
1117 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
1120 dprintk("--> %s entry %p\n", __func__
, entry
);
1123 * If this is just the sequence operation, we did not keep
1124 * a page in the cache entry because we can just use the
1125 * slot info stored in struct nfsd4_sequence that was checked
1126 * against the slot in nfsd4_sequence().
1128 * This occurs when seq->cachethis is FALSE, or when the client
1129 * session inactivity timer fires and a solo sequence operation
1130 * is sent (lease renewal).
1132 if (seq
&& nfsd4_not_cached(resp
)) {
1133 seq
->maxslots
= resp
->cstate
.session
->se_fnumslots
;
1137 if (!nfsd41_copy_replay_data(resp
, entry
)) {
1139 * Not enough room to use the replay rpc header, send the
1140 * cached header. Release all the allocated result pages.
1142 svc_free_res_pages(resp
->rqstp
);
1143 nfsd4_copy_pages(resp
->rqstp
->rq_respages
, entry
->ce_respages
,
1146 /* Release all but the first allocated result page */
1148 resp
->rqstp
->rq_resused
--;
1149 svc_free_res_pages(resp
->rqstp
);
1151 nfsd4_copy_pages(&resp
->rqstp
->rq_respages
[1],
1152 &entry
->ce_respages
[1],
1153 entry
->ce_resused
- 1);
1156 resp
->rqstp
->rq_resused
= entry
->ce_resused
;
1157 resp
->opcnt
= entry
->ce_opcnt
;
1158 resp
->cstate
.iovlen
= entry
->ce_datav
.iov_len
+ entry
->ce_rpchdrlen
;
1159 status
= entry
->ce_status
;
1165 * Set the exchange_id flags returned by the server.
1168 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1170 /* pNFS is not supported */
1171 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1173 /* Referrals are supported, Migration is not. */
1174 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1176 /* set the wire flags to return to client. */
1177 clid
->flags
= new->cl_exchange_flags
;
1181 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1182 struct nfsd4_compound_state
*cstate
,
1183 struct nfsd4_exchange_id
*exid
)
1185 struct nfs4_client
*unconf
, *conf
, *new;
1187 unsigned int strhashval
;
1188 char dname
[HEXDIR_LEN
];
1189 nfs4_verifier verf
= exid
->verifier
;
1190 u32 ip_addr
= svc_addr_in(rqstp
)->sin_addr
.s_addr
;
1192 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1193 " ip_addr=%u flags %x, spa_how %d\n",
1194 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1195 ip_addr
, exid
->flags
, exid
->spa_how
);
1197 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1198 return nfserr_inval
;
1200 /* Currently only support SP4_NONE */
1201 switch (exid
->spa_how
) {
1205 return nfserr_encr_alg_unsupp
;
1207 BUG(); /* checked by xdr code */
1209 return nfserr_serverfault
; /* no excuse :-/ */
1212 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1217 strhashval
= clientstr_hashval(dname
);
1222 conf
= find_confirmed_client_by_str(dname
, strhashval
, true);
1224 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1225 /* 18.35.4 case 8 */
1226 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1227 status
= nfserr_not_same
;
1230 /* Client reboot: destroy old state */
1231 expire_client(conf
);
1234 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1235 /* 18.35.4 case 9 */
1236 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1237 status
= nfserr_perm
;
1240 expire_client(conf
);
1243 if (ip_addr
!= conf
->cl_addr
&&
1244 !(exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
)) {
1245 /* Client collision. 18.35.4 case 3 */
1246 status
= nfserr_clid_inuse
;
1250 * Set bit when the owner id and verifier map to an already
1251 * confirmed client id (18.35.3).
1253 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1256 * Falling into 18.35.4 case 2, possible router replay.
1257 * Leave confirmed record intact and return same result.
1259 copy_verf(conf
, &verf
);
1263 /* 18.35.4 case 7 */
1264 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1265 status
= nfserr_noent
;
1270 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, true);
1273 * Possible retry or client restart. Per 18.35.4 case 4,
1274 * a new unconfirmed record should be generated regardless
1275 * of whether any properties have changed.
1277 expire_client(unconf
);
1282 new = create_client(exid
->clname
, dname
);
1284 status
= nfserr_resource
;
1288 copy_verf(new, &verf
);
1289 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1290 new->cl_addr
= ip_addr
;
1293 add_to_unconfirmed(new, strhashval
);
1295 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1296 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1298 new->cl_slot
.sl_seqid
= 0;
1300 nfsd4_set_ex_flags(new, exid
);
1302 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1303 new->cl_slot
.sl_seqid
, new->cl_exchange_flags
);
1307 nfs4_unlock_state();
1309 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1314 check_slot_seqid(u32 seqid
, struct nfsd4_slot
*slot
)
1316 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__
, seqid
,
1319 /* The slot is in use, and no response has been sent. */
1320 if (slot
->sl_inuse
) {
1321 if (seqid
== slot
->sl_seqid
)
1322 return nfserr_jukebox
;
1324 return nfserr_seq_misordered
;
1327 if (likely(seqid
== slot
->sl_seqid
+ 1))
1330 if (seqid
== slot
->sl_seqid
)
1331 return nfserr_replay_cache
;
1333 if (seqid
== 1 && (slot
->sl_seqid
+ 1) == 0)
1335 /* Misordered replay or misordered new request */
1336 return nfserr_seq_misordered
;
1340 nfsd4_create_session(struct svc_rqst
*rqstp
,
1341 struct nfsd4_compound_state
*cstate
,
1342 struct nfsd4_create_session
*cr_ses
)
1344 u32 ip_addr
= svc_addr_in(rqstp
)->sin_addr
.s_addr
;
1345 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1346 struct nfs4_client
*conf
, *unconf
;
1347 struct nfsd4_slot
*slot
= NULL
;
1351 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1352 conf
= find_confirmed_client(&cr_ses
->clientid
);
1355 slot
= &conf
->cl_slot
;
1356 status
= check_slot_seqid(cr_ses
->seqid
, slot
);
1357 if (status
== nfserr_replay_cache
) {
1358 dprintk("Got a create_session replay! seqid= %d\n",
1360 cstate
->slot
= slot
;
1361 cstate
->status
= status
;
1362 /* Return the cached reply status */
1363 status
= nfsd4_replay_cache_entry(resp
, NULL
);
1365 } else if (cr_ses
->seqid
!= conf
->cl_slot
.sl_seqid
+ 1) {
1366 status
= nfserr_seq_misordered
;
1367 dprintk("Sequence misordered!\n");
1368 dprintk("Expected seqid= %d but got seqid= %d\n",
1369 slot
->sl_seqid
, cr_ses
->seqid
);
1372 conf
->cl_slot
.sl_seqid
++;
1373 } else if (unconf
) {
1374 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1375 (ip_addr
!= unconf
->cl_addr
)) {
1376 status
= nfserr_clid_inuse
;
1380 slot
= &unconf
->cl_slot
;
1381 status
= check_slot_seqid(cr_ses
->seqid
, slot
);
1383 /* an unconfirmed replay returns misordered */
1384 status
= nfserr_seq_misordered
;
1388 slot
->sl_seqid
++; /* from 0 to 1 */
1389 move_to_confirmed(unconf
);
1392 * We do not support RDMA or persistent sessions
1394 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1395 cr_ses
->flags
&= ~SESSION4_RDMA
;
1399 status
= nfserr_stale_clientid
;
1403 status
= alloc_init_session(rqstp
, conf
, cr_ses
);
1407 memcpy(cr_ses
->sessionid
.data
, conf
->cl_sessionid
.data
,
1408 NFS4_MAX_SESSIONID_LEN
);
1409 cr_ses
->seqid
= slot
->sl_seqid
;
1411 slot
->sl_inuse
= true;
1412 cstate
->slot
= slot
;
1413 /* Ensure a page is used for the cache */
1414 slot
->sl_cache_entry
.ce_cachethis
= 1;
1416 nfs4_unlock_state();
1417 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1422 nfsd4_destroy_session(struct svc_rqst
*r
,
1423 struct nfsd4_compound_state
*cstate
,
1424 struct nfsd4_destroy_session
*sessionid
)
1426 struct nfsd4_session
*ses
;
1427 u32 status
= nfserr_badsession
;
1430 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1431 * - Should we return nfserr_back_chan_busy if waiting for
1432 * callbacks on to-be-destroyed session?
1433 * - Do we need to clear any callback info from previous session?
1436 dump_sessionid(__func__
, &sessionid
->sessionid
);
1437 spin_lock(&sessionid_lock
);
1438 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1440 spin_unlock(&sessionid_lock
);
1444 unhash_session(ses
);
1445 spin_unlock(&sessionid_lock
);
1447 /* wait for callbacks */
1448 shutdown_callback_client(ses
->se_client
);
1449 nfsd4_put_session(ses
);
1452 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1457 nfsd4_sequence(struct svc_rqst
*rqstp
,
1458 struct nfsd4_compound_state
*cstate
,
1459 struct nfsd4_sequence
*seq
)
1461 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1462 struct nfsd4_session
*session
;
1463 struct nfsd4_slot
*slot
;
1466 if (resp
->opcnt
!= 1)
1467 return nfserr_sequence_pos
;
1469 spin_lock(&sessionid_lock
);
1470 status
= nfserr_badsession
;
1471 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1475 status
= nfserr_badslot
;
1476 if (seq
->slotid
>= session
->se_fnumslots
)
1479 slot
= &session
->se_slots
[seq
->slotid
];
1480 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1482 status
= check_slot_seqid(seq
->seqid
, slot
);
1483 if (status
== nfserr_replay_cache
) {
1484 cstate
->slot
= slot
;
1485 cstate
->session
= session
;
1486 /* Return the cached reply status and set cstate->status
1487 * for nfsd4_svc_encode_compoundres processing */
1488 status
= nfsd4_replay_cache_entry(resp
, seq
);
1489 cstate
->status
= nfserr_replay_cache
;
1495 /* Success! bump slot seqid */
1496 slot
->sl_inuse
= true;
1497 slot
->sl_seqid
= seq
->seqid
;
1498 slot
->sl_cache_entry
.ce_cachethis
= seq
->cachethis
;
1499 /* Always set the cache entry cachethis for solo sequence */
1500 if (nfsd4_is_solo_sequence(resp
))
1501 slot
->sl_cache_entry
.ce_cachethis
= 1;
1503 cstate
->slot
= slot
;
1504 cstate
->session
= session
;
1507 /* Renew the clientid on success and on replay.
1508 * Hold a session reference until done processing the compound:
1509 * nfsd4_put_session called only if the cstate slot is set.
1511 renew_client(session
->se_client
);
1512 nfsd4_get_session(session
);
1514 spin_unlock(&sessionid_lock
);
1515 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1520 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1521 struct nfsd4_setclientid
*setclid
)
1523 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1524 struct xdr_netobj clname
= {
1525 .len
= setclid
->se_namelen
,
1526 .data
= setclid
->se_name
,
1528 nfs4_verifier clverifier
= setclid
->se_verf
;
1529 unsigned int strhashval
;
1530 struct nfs4_client
*conf
, *unconf
, *new;
1533 char dname
[HEXDIR_LEN
];
1535 if (!check_name(clname
))
1536 return nfserr_inval
;
1538 status
= nfs4_make_rec_clidname(dname
, &clname
);
1543 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1544 * We get here on a DRC miss.
1547 strhashval
= clientstr_hashval(dname
);
1550 conf
= find_confirmed_client_by_str(dname
, strhashval
, false);
1552 /* RFC 3530 14.2.33 CASE 0: */
1553 status
= nfserr_clid_inuse
;
1554 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1555 dprintk("NFSD: setclientid: string in use by client"
1556 " at %pI4\n", &conf
->cl_addr
);
1561 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1562 * has a description of SETCLIENTID request processing consisting
1563 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1565 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, false);
1566 status
= nfserr_resource
;
1569 * RFC 3530 14.2.33 CASE 4:
1570 * placed first, because it is the normal case
1573 expire_client(unconf
);
1574 new = create_client(clname
, dname
);
1578 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1580 * RFC 3530 14.2.33 CASE 1:
1581 * probable callback update
1584 /* Note this is removing unconfirmed {*x***},
1585 * which is stronger than RFC recommended {vxc**}.
1586 * This has the advantage that there is at most
1587 * one {*x***} in either list at any time.
1589 expire_client(unconf
);
1591 new = create_client(clname
, dname
);
1594 copy_clid(new, conf
);
1595 } else if (!unconf
) {
1597 * RFC 3530 14.2.33 CASE 2:
1598 * probable client reboot; state will be removed if
1601 new = create_client(clname
, dname
);
1607 * RFC 3530 14.2.33 CASE 3:
1608 * probable client reboot; state will be removed if
1611 expire_client(unconf
);
1612 new = create_client(clname
, dname
);
1617 copy_verf(new, &clverifier
);
1618 new->cl_addr
= sin
->sin_addr
.s_addr
;
1619 new->cl_flavor
= rqstp
->rq_flavor
;
1620 princ
= svc_gss_principal(rqstp
);
1622 new->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1623 if (new->cl_principal
== NULL
) {
1628 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1630 gen_callback(new, setclid
);
1631 add_to_unconfirmed(new, strhashval
);
1632 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1633 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1634 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1637 nfs4_unlock_state();
1643 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1644 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1645 * bullets, labeled as CASE1 - CASE4 below.
1648 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1649 struct nfsd4_compound_state
*cstate
,
1650 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1652 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1653 struct nfs4_client
*conf
, *unconf
;
1654 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1655 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1658 if (STALE_CLIENTID(clid
))
1659 return nfserr_stale_clientid
;
1661 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1662 * We get here on a DRC miss.
1667 conf
= find_confirmed_client(clid
);
1668 unconf
= find_unconfirmed_client(clid
);
1670 status
= nfserr_clid_inuse
;
1671 if (conf
&& conf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1673 if (unconf
&& unconf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1677 * section 14.2.34 of RFC 3530 has a description of
1678 * SETCLIENTID_CONFIRM request processing consisting
1679 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1681 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
1683 * RFC 3530 14.2.34 CASE 1:
1686 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
1687 status
= nfserr_clid_inuse
;
1689 /* XXX: We just turn off callbacks until we can handle
1690 * change request correctly. */
1691 atomic_set(&conf
->cl_cb_conn
.cb_set
, 0);
1692 expire_client(unconf
);
1696 } else if (conf
&& !unconf
) {
1698 * RFC 3530 14.2.34 CASE 2:
1699 * probable retransmitted request; play it safe and
1702 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
1703 status
= nfserr_clid_inuse
;
1706 } else if (!conf
&& unconf
1707 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
1709 * RFC 3530 14.2.34 CASE 3:
1710 * Normal case; new or rebooted client:
1712 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
1713 status
= nfserr_clid_inuse
;
1716 clientstr_hashval(unconf
->cl_recdir
);
1717 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
1720 nfsd4_remove_clid_dir(conf
);
1721 expire_client(conf
);
1723 move_to_confirmed(unconf
);
1725 nfsd4_probe_callback(conf
);
1728 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
1729 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
1732 * RFC 3530 14.2.34 CASE 4:
1733 * Client probably hasn't noticed that we rebooted yet.
1735 status
= nfserr_stale_clientid
;
1737 /* check that we have hit one of the cases...*/
1738 status
= nfserr_clid_inuse
;
1741 nfs4_unlock_state();
1745 /* OPEN Share state helper functions */
1746 static inline struct nfs4_file
*
1747 alloc_init_file(struct inode
*ino
)
1749 struct nfs4_file
*fp
;
1750 unsigned int hashval
= file_hashval(ino
);
1752 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1754 atomic_set(&fp
->fi_ref
, 1);
1755 INIT_LIST_HEAD(&fp
->fi_hash
);
1756 INIT_LIST_HEAD(&fp
->fi_stateids
);
1757 INIT_LIST_HEAD(&fp
->fi_delegations
);
1758 spin_lock(&recall_lock
);
1759 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1760 spin_unlock(&recall_lock
);
1761 fp
->fi_inode
= igrab(ino
);
1762 fp
->fi_id
= current_fileid
++;
1763 fp
->fi_had_conflict
= false;
1770 nfsd4_free_slab(struct kmem_cache
**slab
)
1774 kmem_cache_destroy(*slab
);
1779 nfsd4_free_slabs(void)
1781 nfsd4_free_slab(&stateowner_slab
);
1782 nfsd4_free_slab(&file_slab
);
1783 nfsd4_free_slab(&stateid_slab
);
1784 nfsd4_free_slab(&deleg_slab
);
1788 nfsd4_init_slabs(void)
1790 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1791 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1792 if (stateowner_slab
== NULL
)
1794 file_slab
= kmem_cache_create("nfsd4_files",
1795 sizeof(struct nfs4_file
), 0, 0, NULL
);
1796 if (file_slab
== NULL
)
1798 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1799 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1800 if (stateid_slab
== NULL
)
1802 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1803 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1804 if (deleg_slab
== NULL
)
1809 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1814 nfs4_free_stateowner(struct kref
*kref
)
1816 struct nfs4_stateowner
*sop
=
1817 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1818 kfree(sop
->so_owner
.data
);
1819 kmem_cache_free(stateowner_slab
, sop
);
1822 static inline struct nfs4_stateowner
*
1823 alloc_stateowner(struct xdr_netobj
*owner
)
1825 struct nfs4_stateowner
*sop
;
1827 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1828 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1829 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1830 sop
->so_owner
.len
= owner
->len
;
1831 kref_init(&sop
->so_ref
);
1834 kmem_cache_free(stateowner_slab
, sop
);
1839 static struct nfs4_stateowner
*
1840 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1841 struct nfs4_stateowner
*sop
;
1842 struct nfs4_replay
*rp
;
1843 unsigned int idhashval
;
1845 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1847 idhashval
= ownerid_hashval(current_ownerid
);
1848 INIT_LIST_HEAD(&sop
->so_idhash
);
1849 INIT_LIST_HEAD(&sop
->so_strhash
);
1850 INIT_LIST_HEAD(&sop
->so_perclient
);
1851 INIT_LIST_HEAD(&sop
->so_stateids
);
1852 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1853 INIT_LIST_HEAD(&sop
->so_close_lru
);
1855 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1856 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1857 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1858 sop
->so_is_open_owner
= 1;
1859 sop
->so_id
= current_ownerid
++;
1860 sop
->so_client
= clp
;
1861 sop
->so_seqid
= open
->op_seqid
;
1862 sop
->so_confirmed
= 0;
1863 rp
= &sop
->so_replay
;
1864 rp
->rp_status
= nfserr_serverfault
;
1866 rp
->rp_buf
= rp
->rp_ibuf
;
1871 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1872 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1873 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1875 INIT_LIST_HEAD(&stp
->st_hash
);
1876 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1877 INIT_LIST_HEAD(&stp
->st_lockowners
);
1878 INIT_LIST_HEAD(&stp
->st_perfile
);
1879 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1880 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1881 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1882 stp
->st_stateowner
= sop
;
1885 stp
->st_stateid
.si_boot
= get_seconds();
1886 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1887 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1888 stp
->st_stateid
.si_generation
= 0;
1889 stp
->st_access_bmap
= 0;
1890 stp
->st_deny_bmap
= 0;
1891 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
1892 &stp
->st_access_bmap
);
1893 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1894 stp
->st_openstp
= NULL
;
1898 move_to_close_lru(struct nfs4_stateowner
*sop
)
1900 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1902 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1903 sop
->so_time
= get_seconds();
1907 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1910 return (sop
->so_owner
.len
== owner
->len
) &&
1911 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1912 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1915 static struct nfs4_stateowner
*
1916 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1918 struct nfs4_stateowner
*so
= NULL
;
1920 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1921 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
1927 /* search file_hashtbl[] for file */
1928 static struct nfs4_file
*
1929 find_file(struct inode
*ino
)
1931 unsigned int hashval
= file_hashval(ino
);
1932 struct nfs4_file
*fp
;
1934 spin_lock(&recall_lock
);
1935 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
1936 if (fp
->fi_inode
== ino
) {
1938 spin_unlock(&recall_lock
);
1942 spin_unlock(&recall_lock
);
1946 static inline int access_valid(u32 x
, u32 minorversion
)
1948 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
1950 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
1952 x
&= ~NFS4_SHARE_ACCESS_MASK
;
1953 if (minorversion
&& x
) {
1954 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
1956 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
1958 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
1965 static inline int deny_valid(u32 x
)
1967 /* Note: unlike access bits, deny bits may be zero. */
1968 return x
<= NFS4_SHARE_DENY_BOTH
;
1972 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1973 * st_{access,deny}_bmap field of the stateid, in order to track not
1974 * only what share bits are currently in force, but also what
1975 * combinations of share bits previous opens have used. This allows us
1976 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1977 * return an error if the client attempt to downgrade to a combination
1978 * of share bits not explicable by closing some of its previous opens.
1980 * XXX: This enforcement is actually incomplete, since we don't keep
1981 * track of access/deny bit combinations; so, e.g., we allow:
1983 * OPEN allow read, deny write
1984 * OPEN allow both, deny none
1985 * DOWNGRADE allow read, deny none
1987 * which we should reject.
1990 set_access(unsigned int *access
, unsigned long bmap
) {
1994 for (i
= 1; i
< 4; i
++) {
1995 if (test_bit(i
, &bmap
))
2001 set_deny(unsigned int *deny
, unsigned long bmap
) {
2005 for (i
= 0; i
< 4; i
++) {
2006 if (test_bit(i
, &bmap
))
2012 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
2013 unsigned int access
, deny
;
2015 set_access(&access
, stp
->st_access_bmap
);
2016 set_deny(&deny
, stp
->st_deny_bmap
);
2017 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
2023 * Called to check deny when READ with all zero stateid or
2024 * WRITE with all zero or all one stateid
2027 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2029 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2030 struct nfs4_file
*fp
;
2031 struct nfs4_stateid
*stp
;
2034 dprintk("NFSD: nfs4_share_conflict\n");
2036 fp
= find_file(ino
);
2039 ret
= nfserr_locked
;
2040 /* Search for conflicting share reservations */
2041 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2042 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2043 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2053 nfs4_file_downgrade(struct file
*filp
, unsigned int share_access
)
2055 if (share_access
& NFS4_SHARE_ACCESS_WRITE
) {
2056 drop_file_write_access(filp
);
2057 filp
->f_mode
= (filp
->f_mode
| FMODE_READ
) & ~FMODE_WRITE
;
2062 * Spawn a thread to perform a recall on the delegation represented
2063 * by the lease (file_lock)
2065 * Called from break_lease() with lock_kernel() held.
2066 * Note: we assume break_lease will only call this *once* for any given
2070 void nfsd_break_deleg_cb(struct file_lock
*fl
)
2072 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2074 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
2078 /* We're assuming the state code never drops its reference
2079 * without first removing the lease. Since we're in this lease
2080 * callback (and since the lease code is serialized by the kernel
2081 * lock) we know the server hasn't removed the lease yet, we know
2082 * it's safe to take a reference: */
2083 atomic_inc(&dp
->dl_count
);
2084 atomic_inc(&dp
->dl_client
->cl_count
);
2086 spin_lock(&recall_lock
);
2087 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2088 spin_unlock(&recall_lock
);
2090 /* only place dl_time is set. protected by lock_kernel*/
2091 dp
->dl_time
= get_seconds();
2094 * We don't want the locks code to timeout the lease for us;
2095 * we'll remove it ourself if the delegation isn't returned
2098 fl
->fl_break_time
= 0;
2100 dp
->dl_file
->fi_had_conflict
= true;
2101 nfsd4_cb_recall(dp
);
2105 * The file_lock is being reapd.
2107 * Called by locks_free_lock() with lock_kernel() held.
2110 void nfsd_release_deleg_cb(struct file_lock
*fl
)
2112 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2114 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
2116 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
2118 dp
->dl_flock
= NULL
;
2122 * Set the delegation file_lock back pointer.
2124 * Called from setlease() with lock_kernel() held.
2127 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
2129 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
2131 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
2138 * Called from setlease() with lock_kernel() held
2141 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
2143 struct nfs4_delegation
*onlistd
=
2144 (struct nfs4_delegation
*)onlist
->fl_owner
;
2145 struct nfs4_delegation
*tryd
=
2146 (struct nfs4_delegation
*)try->fl_owner
;
2148 if (onlist
->fl_lmops
!= try->fl_lmops
)
2151 return onlistd
->dl_client
== tryd
->dl_client
;
2156 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2159 return lease_modify(onlist
, arg
);
2164 static struct lock_manager_operations nfsd_lease_mng_ops
= {
2165 .fl_break
= nfsd_break_deleg_cb
,
2166 .fl_release_private
= nfsd_release_deleg_cb
,
2167 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
2168 .fl_mylease
= nfsd_same_client_deleg_cb
,
2169 .fl_change
= nfsd_change_deleg_cb
,
2174 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2175 struct nfsd4_open
*open
)
2177 clientid_t
*clientid
= &open
->op_clientid
;
2178 struct nfs4_client
*clp
= NULL
;
2179 unsigned int strhashval
;
2180 struct nfs4_stateowner
*sop
= NULL
;
2182 if (!check_name(open
->op_owner
))
2183 return nfserr_inval
;
2185 if (STALE_CLIENTID(&open
->op_clientid
))
2186 return nfserr_stale_clientid
;
2188 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
2189 sop
= find_openstateowner_str(strhashval
, open
);
2190 open
->op_stateowner
= sop
;
2192 /* Make sure the client's lease hasn't expired. */
2193 clp
= find_confirmed_client(clientid
);
2195 return nfserr_expired
;
2198 /* When sessions are used, skip open sequenceid processing */
2199 if (nfsd4_has_session(cstate
))
2201 if (!sop
->so_confirmed
) {
2202 /* Replace unconfirmed owners without checking for replay. */
2203 clp
= sop
->so_client
;
2204 release_openowner(sop
);
2205 open
->op_stateowner
= NULL
;
2208 if (open
->op_seqid
== sop
->so_seqid
- 1) {
2209 if (sop
->so_replay
.rp_buflen
)
2210 return nfserr_replay_me
;
2211 /* The original OPEN failed so spectacularly
2212 * that we don't even have replay data saved!
2213 * Therefore, we have no choice but to continue
2214 * processing this OPEN; presumably, we'll
2215 * fail again for the same reason.
2217 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2220 if (open
->op_seqid
!= sop
->so_seqid
)
2221 return nfserr_bad_seqid
;
2223 if (open
->op_stateowner
== NULL
) {
2224 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2226 return nfserr_resource
;
2227 open
->op_stateowner
= sop
;
2229 list_del_init(&sop
->so_close_lru
);
2230 renew_client(sop
->so_client
);
2234 static inline __be32
2235 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2237 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2238 return nfserr_openmode
;
2243 static struct nfs4_delegation
*
2244 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2246 struct nfs4_delegation
*dp
;
2248 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
2249 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
2256 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2257 struct nfs4_delegation
**dp
)
2260 __be32 status
= nfserr_bad_stateid
;
2262 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2265 flags
= open
->op_share_access
== NFS4_SHARE_ACCESS_READ
?
2266 RD_STATE
: WR_STATE
;
2267 status
= nfs4_check_delegmode(*dp
, flags
);
2271 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2275 open
->op_stateowner
->so_confirmed
= 1;
2280 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2282 struct nfs4_stateid
*local
;
2283 __be32 status
= nfserr_share_denied
;
2284 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2286 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2287 /* ignore lock owners */
2288 if (local
->st_stateowner
->so_is_open_owner
== 0)
2290 /* remember if we have seen this open owner */
2291 if (local
->st_stateowner
== sop
)
2293 /* check for conflicting share reservations */
2294 if (!test_share(local
, open
))
2302 static inline struct nfs4_stateid
*
2303 nfs4_alloc_stateid(void)
2305 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2309 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2310 struct nfs4_delegation
*dp
,
2311 struct svc_fh
*cur_fh
, int flags
)
2313 struct nfs4_stateid
*stp
;
2315 stp
= nfs4_alloc_stateid();
2317 return nfserr_resource
;
2320 get_file(dp
->dl_vfs_file
);
2321 stp
->st_vfs_file
= dp
->dl_vfs_file
;
2324 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, flags
,
2327 if (status
== nfserr_dropit
)
2328 status
= nfserr_jukebox
;
2329 kmem_cache_free(stateid_slab
, stp
);
2337 static inline __be32
2338 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2339 struct nfsd4_open
*open
)
2341 struct iattr iattr
= {
2342 .ia_valid
= ATTR_SIZE
,
2345 if (!open
->op_truncate
)
2347 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2348 return nfserr_inval
;
2349 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2353 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2355 struct file
*filp
= stp
->st_vfs_file
;
2356 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
2357 unsigned int share_access
, new_writer
;
2360 set_access(&share_access
, stp
->st_access_bmap
);
2361 new_writer
= (~share_access
) & open
->op_share_access
2362 & NFS4_SHARE_ACCESS_WRITE
;
2365 int err
= get_write_access(inode
);
2367 return nfserrno(err
);
2368 err
= mnt_want_write(cur_fh
->fh_export
->ex_path
.mnt
);
2370 return nfserrno(err
);
2371 file_take_write(filp
);
2373 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2376 put_write_access(inode
);
2379 /* remember the open */
2380 filp
->f_mode
|= open
->op_share_access
;
2381 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
2382 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2389 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2391 open
->op_stateowner
->so_confirmed
= 1;
2392 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2396 * Attempt to hand out a delegation.
2399 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2401 struct nfs4_delegation
*dp
;
2402 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2403 struct nfs4_cb_conn
*cb
= &sop
->so_client
->cl_cb_conn
;
2404 struct file_lock fl
, *flp
= &fl
;
2405 int status
, flag
= 0;
2407 flag
= NFS4_OPEN_DELEGATE_NONE
;
2408 open
->op_recall
= 0;
2409 switch (open
->op_claim_type
) {
2410 case NFS4_OPEN_CLAIM_PREVIOUS
:
2411 if (!atomic_read(&cb
->cb_set
))
2412 open
->op_recall
= 1;
2413 flag
= open
->op_delegate_type
;
2414 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2417 case NFS4_OPEN_CLAIM_NULL
:
2418 /* Let's not give out any delegations till everyone's
2419 * had the chance to reclaim theirs.... */
2420 if (locks_in_grace())
2422 if (!atomic_read(&cb
->cb_set
) || !sop
->so_confirmed
)
2424 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2425 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2427 flag
= NFS4_OPEN_DELEGATE_READ
;
2433 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2435 flag
= NFS4_OPEN_DELEGATE_NONE
;
2438 locks_init_lock(&fl
);
2439 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
2440 fl
.fl_flags
= FL_LEASE
;
2441 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2442 fl
.fl_end
= OFFSET_MAX
;
2443 fl
.fl_owner
= (fl_owner_t
)dp
;
2444 fl
.fl_file
= stp
->st_vfs_file
;
2445 fl
.fl_pid
= current
->tgid
;
2447 /* vfs_setlease checks to see if delegation should be handed out.
2448 * the lock_manager callbacks fl_mylease and fl_change are used
2450 if ((status
= vfs_setlease(stp
->st_vfs_file
, fl
.fl_type
, &flp
))) {
2451 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
2452 unhash_delegation(dp
);
2453 flag
= NFS4_OPEN_DELEGATE_NONE
;
2457 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2459 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2460 dp
->dl_stateid
.si_boot
,
2461 dp
->dl_stateid
.si_stateownerid
,
2462 dp
->dl_stateid
.si_fileid
,
2463 dp
->dl_stateid
.si_generation
);
2465 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2466 && flag
== NFS4_OPEN_DELEGATE_NONE
2467 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2468 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2469 open
->op_delegate_type
= flag
;
2473 * called with nfs4_lock_state() held.
2476 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2478 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2479 struct nfs4_file
*fp
= NULL
;
2480 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2481 struct nfs4_stateid
*stp
= NULL
;
2482 struct nfs4_delegation
*dp
= NULL
;
2485 status
= nfserr_inval
;
2486 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2487 || !deny_valid(open
->op_share_deny
))
2490 * Lookup file; if found, lookup stateid and check open request,
2491 * and check for delegations in the process of being recalled.
2492 * If not found, create the nfs4_file struct
2494 fp
= find_file(ino
);
2496 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2498 status
= nfs4_check_deleg(fp
, open
, &dp
);
2502 status
= nfserr_bad_stateid
;
2503 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2505 status
= nfserr_resource
;
2506 fp
= alloc_init_file(ino
);
2512 * OPEN the file, or upgrade an existing OPEN.
2513 * If truncate fails, the OPEN fails.
2516 /* Stateid was found, this is an OPEN upgrade */
2517 status
= nfs4_upgrade_open(rqstp
, current_fh
, stp
, open
);
2520 update_stateid(&stp
->st_stateid
);
2522 /* Stateid was not found, this is a new OPEN */
2524 if (open
->op_share_access
& NFS4_SHARE_ACCESS_READ
)
2525 flags
|= NFSD_MAY_READ
;
2526 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2527 flags
|= NFSD_MAY_WRITE
;
2528 status
= nfs4_new_open(rqstp
, &stp
, dp
, current_fh
, flags
);
2531 init_stateid(stp
, fp
, open
);
2532 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2534 release_open_stateid(stp
);
2537 if (nfsd4_has_session(&resp
->cstate
))
2538 update_stateid(&stp
->st_stateid
);
2540 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2542 if (nfsd4_has_session(&resp
->cstate
))
2543 open
->op_stateowner
->so_confirmed
= 1;
2546 * Attempt to hand out a delegation. No error return, because the
2547 * OPEN succeeds even if we fail.
2549 nfs4_open_delegation(current_fh
, open
, stp
);
2553 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2554 stp
->st_stateid
.si_boot
, stp
->st_stateid
.si_stateownerid
,
2555 stp
->st_stateid
.si_fileid
, stp
->st_stateid
.si_generation
);
2559 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2560 nfs4_set_claim_prev(open
);
2562 * To finish the open response, we just need to set the rflags.
2564 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2565 if (!open
->op_stateowner
->so_confirmed
&&
2566 !nfsd4_has_session(&resp
->cstate
))
2567 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2573 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2576 struct nfs4_client
*clp
;
2580 dprintk("process_renew(%08x/%08x): starting\n",
2581 clid
->cl_boot
, clid
->cl_id
);
2582 status
= nfserr_stale_clientid
;
2583 if (STALE_CLIENTID(clid
))
2585 clp
= find_confirmed_client(clid
);
2586 status
= nfserr_expired
;
2588 /* We assume the client took too long to RENEW. */
2589 dprintk("nfsd4_renew: clientid not found!\n");
2593 status
= nfserr_cb_path_down
;
2594 if (!list_empty(&clp
->cl_delegations
)
2595 && !atomic_read(&clp
->cl_cb_conn
.cb_set
))
2599 nfs4_unlock_state();
2603 struct lock_manager nfsd4_manager
= {
2607 nfsd4_end_grace(void)
2609 dprintk("NFSD: end of grace period\n");
2610 nfsd4_recdir_purge_old();
2611 locks_end_grace(&nfsd4_manager
);
2615 nfs4_laundromat(void)
2617 struct nfs4_client
*clp
;
2618 struct nfs4_stateowner
*sop
;
2619 struct nfs4_delegation
*dp
;
2620 struct list_head
*pos
, *next
, reaplist
;
2621 time_t cutoff
= get_seconds() - NFSD_LEASE_TIME
;
2622 time_t t
, clientid_val
= NFSD_LEASE_TIME
;
2623 time_t u
, test_val
= NFSD_LEASE_TIME
;
2627 dprintk("NFSD: laundromat service - starting\n");
2628 if (locks_in_grace())
2630 list_for_each_safe(pos
, next
, &client_lru
) {
2631 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2632 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2633 t
= clp
->cl_time
- cutoff
;
2634 if (clientid_val
> t
)
2638 dprintk("NFSD: purging unused client (clientid %08x)\n",
2639 clp
->cl_clientid
.cl_id
);
2640 nfsd4_remove_clid_dir(clp
);
2643 INIT_LIST_HEAD(&reaplist
);
2644 spin_lock(&recall_lock
);
2645 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2646 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2647 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2648 u
= dp
->dl_time
- cutoff
;
2653 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2655 list_move(&dp
->dl_recall_lru
, &reaplist
);
2657 spin_unlock(&recall_lock
);
2658 list_for_each_safe(pos
, next
, &reaplist
) {
2659 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2660 list_del_init(&dp
->dl_recall_lru
);
2661 unhash_delegation(dp
);
2663 test_val
= NFSD_LEASE_TIME
;
2664 list_for_each_safe(pos
, next
, &close_lru
) {
2665 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2666 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2667 u
= sop
->so_time
- cutoff
;
2672 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2674 release_openowner(sop
);
2676 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2677 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2678 nfs4_unlock_state();
2679 return clientid_val
;
2682 static struct workqueue_struct
*laundry_wq
;
2683 static void laundromat_main(struct work_struct
*);
2684 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
2687 laundromat_main(struct work_struct
*not_used
)
2691 t
= nfs4_laundromat();
2692 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
2693 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
2696 static struct nfs4_stateowner
*
2697 search_close_lru(u32 st_id
, int flags
)
2699 struct nfs4_stateowner
*local
= NULL
;
2701 if (flags
& CLOSE_STATE
) {
2702 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
2703 if (local
->so_id
== st_id
)
2711 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
2713 return fhp
->fh_dentry
->d_inode
!= stp
->st_vfs_file
->f_path
.dentry
->d_inode
;
2717 STALE_STATEID(stateid_t
*stateid
)
2719 if (time_after((unsigned long)boot_time
,
2720 (unsigned long)stateid
->si_boot
)) {
2721 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2722 stateid
->si_boot
, stateid
->si_stateownerid
,
2723 stateid
->si_fileid
, stateid
->si_generation
);
2730 EXPIRED_STATEID(stateid_t
*stateid
)
2732 if (time_before((unsigned long)boot_time
,
2733 ((unsigned long)stateid
->si_boot
)) &&
2734 time_before((unsigned long)(stateid
->si_boot
+ lease_time
), get_seconds())) {
2735 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2736 stateid
->si_boot
, stateid
->si_stateownerid
,
2737 stateid
->si_fileid
, stateid
->si_generation
);
2744 stateid_error_map(stateid_t
*stateid
)
2746 if (STALE_STATEID(stateid
))
2747 return nfserr_stale_stateid
;
2748 if (EXPIRED_STATEID(stateid
))
2749 return nfserr_expired
;
2751 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2752 stateid
->si_boot
, stateid
->si_stateownerid
,
2753 stateid
->si_fileid
, stateid
->si_generation
);
2754 return nfserr_bad_stateid
;
2758 access_permit_read(unsigned long access_bmap
)
2760 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
2761 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
2762 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
2766 access_permit_write(unsigned long access_bmap
)
2768 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
2769 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
2773 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
2775 __be32 status
= nfserr_openmode
;
2777 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
2779 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2786 static inline __be32
2787 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2789 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2791 else if (locks_in_grace()) {
2792 /* Answer in remaining cases depends on existance of
2793 * conflicting state; so we must wait out the grace period. */
2794 return nfserr_grace
;
2795 } else if (flags
& WR_STATE
)
2796 return nfs4_share_conflict(current_fh
,
2797 NFS4_SHARE_DENY_WRITE
);
2798 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2799 return nfs4_share_conflict(current_fh
,
2800 NFS4_SHARE_DENY_READ
);
2804 * Allow READ/WRITE during grace period on recovered state only for files
2805 * that are not able to provide mandatory locking.
2808 grace_disallows_io(struct inode
*inode
)
2810 return locks_in_grace() && mandatory_lock(inode
);
2813 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, int flags
)
2816 * When sessions are used the stateid generation number is ignored
2819 if ((flags
& HAS_SESSION
) && in
->si_generation
== 0)
2822 /* If the client sends us a stateid from the future, it's buggy: */
2823 if (in
->si_generation
> ref
->si_generation
)
2824 return nfserr_bad_stateid
;
2826 * The following, however, can happen. For example, if the
2827 * client sends an open and some IO at the same time, the open
2828 * may bump si_generation while the IO is still in flight.
2829 * Thanks to hard links and renames, the client never knows what
2830 * file an open will affect. So it could avoid that situation
2831 * only by serializing all opens and IO from the same open
2832 * owner. To recover from the old_stateid error, the client
2833 * will just have to retry the IO:
2835 if (in
->si_generation
< ref
->si_generation
)
2836 return nfserr_old_stateid
;
2841 static int is_delegation_stateid(stateid_t
*stateid
)
2843 return stateid
->si_fileid
== 0;
2847 * Checks for stateid operations
2850 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
2851 stateid_t
*stateid
, int flags
, struct file
**filpp
)
2853 struct nfs4_stateid
*stp
= NULL
;
2854 struct nfs4_delegation
*dp
= NULL
;
2855 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2856 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2862 if (grace_disallows_io(ino
))
2863 return nfserr_grace
;
2865 if (nfsd4_has_session(cstate
))
2866 flags
|= HAS_SESSION
;
2868 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2869 return check_special_stateids(current_fh
, stateid
, flags
);
2871 status
= nfserr_stale_stateid
;
2872 if (STALE_STATEID(stateid
))
2875 status
= nfserr_bad_stateid
;
2876 if (is_delegation_stateid(stateid
)) {
2877 dp
= find_delegation_stateid(ino
, stateid
);
2879 status
= stateid_error_map(stateid
);
2882 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
,
2886 status
= nfs4_check_delegmode(dp
, flags
);
2889 renew_client(dp
->dl_client
);
2891 *filpp
= dp
->dl_vfs_file
;
2892 } else { /* open or lock stateid */
2893 stp
= find_stateid(stateid
, flags
);
2895 status
= stateid_error_map(stateid
);
2898 if (nfs4_check_fh(current_fh
, stp
))
2900 if (!stp
->st_stateowner
->so_confirmed
)
2902 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
2906 status
= nfs4_check_openmode(stp
, flags
);
2909 renew_client(stp
->st_stateowner
->so_client
);
2911 *filpp
= stp
->st_vfs_file
;
2921 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2922 RD_STATE
: WR_STATE
;
2926 * Checks for sequence id mutating operations.
2929 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
2930 stateid_t
*stateid
, int flags
,
2931 struct nfs4_stateowner
**sopp
,
2932 struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
2934 struct nfs4_stateid
*stp
;
2935 struct nfs4_stateowner
*sop
;
2936 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2939 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2940 "stateid = (%08x/%08x/%08x/%08x)\n", seqid
,
2941 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2942 stateid
->si_generation
);
2947 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
2948 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2949 return nfserr_bad_stateid
;
2952 if (STALE_STATEID(stateid
))
2953 return nfserr_stale_stateid
;
2955 if (nfsd4_has_session(cstate
))
2956 flags
|= HAS_SESSION
;
2959 * We return BAD_STATEID if filehandle doesn't match stateid,
2960 * the confirmed flag is incorrecly set, or the generation
2961 * number is incorrect.
2963 stp
= find_stateid(stateid
, flags
);
2966 * Also, we should make sure this isn't just the result of
2969 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
2971 return stateid_error_map(stateid
);
2977 *sopp
= sop
= stp
->st_stateowner
;
2980 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
2981 struct nfs4_client
*clp
= sop
->so_client
;
2985 lkflg
= setlkflg(lock
->lk_type
);
2987 if (lock
->lk_is_new
) {
2988 if (!sop
->so_is_open_owner
)
2989 return nfserr_bad_stateid
;
2990 if (!(flags
& HAS_SESSION
) &&
2991 !same_clid(&clp
->cl_clientid
, lockclid
))
2992 return nfserr_bad_stateid
;
2993 /* stp is the open stateid */
2994 status
= nfs4_check_openmode(stp
, lkflg
);
2998 /* stp is the lock stateid */
2999 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
3005 if (nfs4_check_fh(current_fh
, stp
)) {
3006 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3007 return nfserr_bad_stateid
;
3011 * We now validate the seqid and stateid generation numbers.
3012 * For the moment, we ignore the possibility of
3013 * generation number wraparound.
3015 if (!(flags
& HAS_SESSION
) && seqid
!= sop
->so_seqid
)
3018 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
3019 dprintk("NFSD: preprocess_seqid_op: expected"
3020 " unconfirmed stateowner!\n");
3021 return nfserr_bad_stateid
;
3023 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
3024 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3025 " confirmed yet!\n");
3026 return nfserr_bad_stateid
;
3028 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, flags
);
3031 renew_client(sop
->so_client
);
3035 if (seqid
== sop
->so_seqid
- 1) {
3036 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3037 /* indicate replay to calling function */
3038 return nfserr_replay_me
;
3040 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3041 sop
->so_seqid
, seqid
);
3043 return nfserr_bad_seqid
;
3047 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3048 struct nfsd4_open_confirm
*oc
)
3051 struct nfs4_stateowner
*sop
;
3052 struct nfs4_stateid
*stp
;
3054 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3055 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3056 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3058 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3064 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3065 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3066 CONFIRM
| OPEN_STATE
,
3067 &oc
->oc_stateowner
, &stp
, NULL
)))
3070 sop
= oc
->oc_stateowner
;
3071 sop
->so_confirmed
= 1;
3072 update_stateid(&stp
->st_stateid
);
3073 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3074 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3075 "stateid=(%08x/%08x/%08x/%08x)\n", oc
->oc_seqid
,
3076 stp
->st_stateid
.si_boot
,
3077 stp
->st_stateid
.si_stateownerid
,
3078 stp
->st_stateid
.si_fileid
,
3079 stp
->st_stateid
.si_generation
);
3081 nfsd4_create_clid_dir(sop
->so_client
);
3083 if (oc
->oc_stateowner
) {
3084 nfs4_get_stateowner(oc
->oc_stateowner
);
3085 cstate
->replay_owner
= oc
->oc_stateowner
;
3087 nfs4_unlock_state();
3093 * unset all bits in union bitmap (bmap) that
3094 * do not exist in share (from successful OPEN_DOWNGRADE)
3097 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
3100 for (i
= 1; i
< 4; i
++) {
3101 if ((i
& access
) != i
)
3102 __clear_bit(i
, bmap
);
3107 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3110 for (i
= 0; i
< 4; i
++) {
3111 if ((i
& deny
) != i
)
3112 __clear_bit(i
, bmap
);
3117 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3118 struct nfsd4_compound_state
*cstate
,
3119 struct nfsd4_open_downgrade
*od
)
3122 struct nfs4_stateid
*stp
;
3123 unsigned int share_access
;
3125 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3126 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3127 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3129 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3130 || !deny_valid(od
->od_share_deny
))
3131 return nfserr_inval
;
3134 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3138 &od
->od_stateowner
, &stp
, NULL
)))
3141 status
= nfserr_inval
;
3142 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3143 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3144 stp
->st_access_bmap
, od
->od_share_access
);
3147 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3148 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3149 stp
->st_deny_bmap
, od
->od_share_deny
);
3152 set_access(&share_access
, stp
->st_access_bmap
);
3153 nfs4_file_downgrade(stp
->st_vfs_file
,
3154 share_access
& ~od
->od_share_access
);
3156 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
3157 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3159 update_stateid(&stp
->st_stateid
);
3160 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3163 if (od
->od_stateowner
) {
3164 nfs4_get_stateowner(od
->od_stateowner
);
3165 cstate
->replay_owner
= od
->od_stateowner
;
3167 nfs4_unlock_state();
3172 * nfs4_unlock_state() called after encode
3175 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3176 struct nfsd4_close
*close
)
3179 struct nfs4_stateid
*stp
;
3181 dprintk("NFSD: nfsd4_close on file %.*s\n",
3182 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3183 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3186 /* check close_lru for replay */
3187 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3190 OPEN_STATE
| CLOSE_STATE
,
3191 &close
->cl_stateowner
, &stp
, NULL
)))
3194 update_stateid(&stp
->st_stateid
);
3195 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3197 /* release_stateid() calls nfsd_close() if needed */
3198 release_open_stateid(stp
);
3200 /* place unused nfs4_stateowners on so_close_lru list to be
3201 * released by the laundromat service after the lease period
3202 * to enable us to handle CLOSE replay
3204 if (list_empty(&close
->cl_stateowner
->so_stateids
))
3205 move_to_close_lru(close
->cl_stateowner
);
3207 if (close
->cl_stateowner
) {
3208 nfs4_get_stateowner(close
->cl_stateowner
);
3209 cstate
->replay_owner
= close
->cl_stateowner
;
3211 nfs4_unlock_state();
3216 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3217 struct nfsd4_delegreturn
*dr
)
3219 struct nfs4_delegation
*dp
;
3220 stateid_t
*stateid
= &dr
->dr_stateid
;
3221 struct inode
*inode
;
3225 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3227 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3229 if (nfsd4_has_session(cstate
))
3230 flags
|= HAS_SESSION
;
3232 status
= nfserr_bad_stateid
;
3233 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3235 status
= nfserr_stale_stateid
;
3236 if (STALE_STATEID(stateid
))
3238 status
= nfserr_bad_stateid
;
3239 if (!is_delegation_stateid(stateid
))
3241 dp
= find_delegation_stateid(inode
, stateid
);
3243 status
= stateid_error_map(stateid
);
3246 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, flags
);
3249 renew_client(dp
->dl_client
);
3251 unhash_delegation(dp
);
3253 nfs4_unlock_state();
3260 * Lock owner state (byte-range locks)
3262 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3263 #define LOCK_HASH_BITS 8
3264 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3265 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3268 end_offset(u64 start
, u64 len
)
3273 return end
>= start
? end
: NFS4_MAX_UINT64
;
3276 /* last octet in a range */
3278 last_byte_offset(u64 start
, u64 len
)
3284 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3287 #define lockownerid_hashval(id) \
3288 ((id) & LOCK_HASH_MASK)
3290 static inline unsigned int
3291 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3292 struct xdr_netobj
*ownername
)
3294 return (file_hashval(inode
) + cl_id
3295 + opaque_hashval(ownername
->data
, ownername
->len
))
3299 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3300 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3301 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
3303 static struct nfs4_stateid
*
3304 find_stateid(stateid_t
*stid
, int flags
)
3306 struct nfs4_stateid
*local
;
3307 u32 st_id
= stid
->si_stateownerid
;
3308 u32 f_id
= stid
->si_fileid
;
3309 unsigned int hashval
;
3311 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
3312 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
3313 hashval
= stateid_hashval(st_id
, f_id
);
3314 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
3315 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3316 (local
->st_stateid
.si_fileid
== f_id
))
3321 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
3322 hashval
= stateid_hashval(st_id
, f_id
);
3323 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
3324 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3325 (local
->st_stateid
.si_fileid
== f_id
))
3332 static struct nfs4_delegation
*
3333 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3335 struct nfs4_file
*fp
;
3336 struct nfs4_delegation
*dl
;
3338 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3339 stid
->si_boot
, stid
->si_stateownerid
,
3340 stid
->si_fileid
, stid
->si_generation
);
3342 fp
= find_file(ino
);
3345 dl
= find_delegation_file(fp
, stid
);
3351 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3352 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3353 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3354 * locking, this prevents us from being completely protocol-compliant. The
3355 * real solution to this problem is to start using unsigned file offsets in
3356 * the VFS, but this is a very deep change!
3359 nfs4_transform_lock_offset(struct file_lock
*lock
)
3361 if (lock
->fl_start
< 0)
3362 lock
->fl_start
= OFFSET_MAX
;
3363 if (lock
->fl_end
< 0)
3364 lock
->fl_end
= OFFSET_MAX
;
3367 /* Hack!: For now, we're defining this just so we can use a pointer to it
3368 * as a unique cookie to identify our (NFSv4's) posix locks. */
3369 static struct lock_manager_operations nfsd_posix_mng_ops
= {
3373 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3375 struct nfs4_stateowner
*sop
;
3378 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3379 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3380 hval
= lockownerid_hashval(sop
->so_id
);
3381 kref_get(&sop
->so_ref
);
3383 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3385 deny
->ld_sop
= NULL
;
3386 deny
->ld_clientid
.cl_boot
= 0;
3387 deny
->ld_clientid
.cl_id
= 0;
3389 deny
->ld_start
= fl
->fl_start
;
3390 deny
->ld_length
= NFS4_MAX_UINT64
;
3391 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3392 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3393 deny
->ld_type
= NFS4_READ_LT
;
3394 if (fl
->fl_type
!= F_RDLCK
)
3395 deny
->ld_type
= NFS4_WRITE_LT
;
3398 static struct nfs4_stateowner
*
3399 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3400 struct xdr_netobj
*owner
)
3402 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3403 struct nfs4_stateowner
*op
;
3405 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3406 if (same_owner_str(op
, owner
, clid
))
3413 * Alloc a lock owner structure.
3414 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3417 * strhashval = lock_ownerstr_hashval
3420 static struct nfs4_stateowner
*
3421 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3422 struct nfs4_stateowner
*sop
;
3423 struct nfs4_replay
*rp
;
3424 unsigned int idhashval
;
3426 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3428 idhashval
= lockownerid_hashval(current_ownerid
);
3429 INIT_LIST_HEAD(&sop
->so_idhash
);
3430 INIT_LIST_HEAD(&sop
->so_strhash
);
3431 INIT_LIST_HEAD(&sop
->so_perclient
);
3432 INIT_LIST_HEAD(&sop
->so_stateids
);
3433 INIT_LIST_HEAD(&sop
->so_perstateid
);
3434 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3436 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3437 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3438 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3439 sop
->so_is_open_owner
= 0;
3440 sop
->so_id
= current_ownerid
++;
3441 sop
->so_client
= clp
;
3442 /* It is the openowner seqid that will be incremented in encode in the
3443 * case of new lockowners; so increment the lock seqid manually: */
3444 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3445 sop
->so_confirmed
= 1;
3446 rp
= &sop
->so_replay
;
3447 rp
->rp_status
= nfserr_serverfault
;
3449 rp
->rp_buf
= rp
->rp_ibuf
;
3453 static struct nfs4_stateid
*
3454 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3456 struct nfs4_stateid
*stp
;
3457 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3459 stp
= nfs4_alloc_stateid();
3462 INIT_LIST_HEAD(&stp
->st_hash
);
3463 INIT_LIST_HEAD(&stp
->st_perfile
);
3464 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3465 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3466 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3467 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3468 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3469 stp
->st_stateowner
= sop
;
3472 stp
->st_stateid
.si_boot
= get_seconds();
3473 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3474 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3475 stp
->st_stateid
.si_generation
= 0;
3476 stp
->st_vfs_file
= open_stp
->st_vfs_file
; /* FIXME refcount?? */
3477 stp
->st_access_bmap
= open_stp
->st_access_bmap
;
3478 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3479 stp
->st_openstp
= open_stp
;
3486 check_lock_length(u64 offset
, u64 length
)
3488 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3489 LOFF_OVERFLOW(offset
, length
)));
3496 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3497 struct nfsd4_lock
*lock
)
3499 struct nfs4_stateowner
*open_sop
= NULL
;
3500 struct nfs4_stateowner
*lock_sop
= NULL
;
3501 struct nfs4_stateid
*lock_stp
;
3503 struct file_lock file_lock
;
3504 struct file_lock conflock
;
3506 unsigned int strhashval
;
3510 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3511 (long long) lock
->lk_offset
,
3512 (long long) lock
->lk_length
);
3514 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3515 return nfserr_inval
;
3517 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3518 S_IFREG
, NFSD_MAY_LOCK
))) {
3519 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3525 if (lock
->lk_is_new
) {
3527 * Client indicates that this is a new lockowner.
3528 * Use open owner and open stateid to create lock owner and
3531 struct nfs4_stateid
*open_stp
= NULL
;
3532 struct nfs4_file
*fp
;
3534 status
= nfserr_stale_clientid
;
3535 if (!nfsd4_has_session(cstate
) &&
3536 STALE_CLIENTID(&lock
->lk_new_clientid
))
3539 /* validate and update open stateid and open seqid */
3540 status
= nfs4_preprocess_seqid_op(cstate
,
3541 lock
->lk_new_open_seqid
,
3542 &lock
->lk_new_open_stateid
,
3544 &lock
->lk_replay_owner
, &open_stp
,
3548 open_sop
= lock
->lk_replay_owner
;
3549 /* create lockowner and lock stateid */
3550 fp
= open_stp
->st_file
;
3551 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3552 open_sop
->so_client
->cl_clientid
.cl_id
,
3553 &lock
->v
.new.owner
);
3554 /* XXX: Do we need to check for duplicate stateowners on
3555 * the same file, or should they just be allowed (and
3556 * create new stateids)? */
3557 status
= nfserr_resource
;
3558 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3559 open_sop
->so_client
, open_stp
, lock
);
3560 if (lock_sop
== NULL
)
3562 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3563 if (lock_stp
== NULL
)
3566 /* lock (lock owner + lock stateid) already exists */
3567 status
= nfs4_preprocess_seqid_op(cstate
,
3568 lock
->lk_old_lock_seqid
,
3569 &lock
->lk_old_lock_stateid
,
3571 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3574 lock_sop
= lock
->lk_replay_owner
;
3576 /* lock->lk_replay_owner and lock_stp have been created or found */
3577 filp
= lock_stp
->st_vfs_file
;
3579 status
= nfserr_grace
;
3580 if (locks_in_grace() && !lock
->lk_reclaim
)
3582 status
= nfserr_no_grace
;
3583 if (!locks_in_grace() && lock
->lk_reclaim
)
3586 locks_init_lock(&file_lock
);
3587 switch (lock
->lk_type
) {
3590 file_lock
.fl_type
= F_RDLCK
;
3594 case NFS4_WRITEW_LT
:
3595 file_lock
.fl_type
= F_WRLCK
;
3599 status
= nfserr_inval
;
3602 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3603 file_lock
.fl_pid
= current
->tgid
;
3604 file_lock
.fl_file
= filp
;
3605 file_lock
.fl_flags
= FL_POSIX
;
3606 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3608 file_lock
.fl_start
= lock
->lk_offset
;
3609 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3610 nfs4_transform_lock_offset(&file_lock
);
3613 * Try to lock the file in the VFS.
3614 * Note: locks.c uses the BKL to protect the inode's lock list.
3617 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
3619 case 0: /* success! */
3620 update_stateid(&lock_stp
->st_stateid
);
3621 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3625 case (EAGAIN
): /* conflock holds conflicting lock */
3626 status
= nfserr_denied
;
3627 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3628 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3631 status
= nfserr_deadlock
;
3634 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3635 status
= nfserr_resource
;
3639 if (status
&& lock
->lk_is_new
&& lock_sop
)
3640 release_lockowner(lock_sop
);
3641 if (lock
->lk_replay_owner
) {
3642 nfs4_get_stateowner(lock
->lk_replay_owner
);
3643 cstate
->replay_owner
= lock
->lk_replay_owner
;
3645 nfs4_unlock_state();
3650 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3651 * so we do a temporary open here just to get an open file to pass to
3652 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3655 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3660 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3663 err
= vfs_test_lock(file
, lock
);
3672 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3673 struct nfsd4_lockt
*lockt
)
3675 struct inode
*inode
;
3676 struct file_lock file_lock
;
3680 if (locks_in_grace())
3681 return nfserr_grace
;
3683 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3684 return nfserr_inval
;
3686 lockt
->lt_stateowner
= NULL
;
3689 status
= nfserr_stale_clientid
;
3690 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
3693 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3694 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3695 if (status
== nfserr_symlink
)
3696 status
= nfserr_inval
;
3700 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3701 locks_init_lock(&file_lock
);
3702 switch (lockt
->lt_type
) {
3705 file_lock
.fl_type
= F_RDLCK
;
3708 case NFS4_WRITEW_LT
:
3709 file_lock
.fl_type
= F_WRLCK
;
3712 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3713 status
= nfserr_inval
;
3717 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
3718 &lockt
->lt_clientid
, &lockt
->lt_owner
);
3719 if (lockt
->lt_stateowner
)
3720 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
3721 file_lock
.fl_pid
= current
->tgid
;
3722 file_lock
.fl_flags
= FL_POSIX
;
3724 file_lock
.fl_start
= lockt
->lt_offset
;
3725 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
3727 nfs4_transform_lock_offset(&file_lock
);
3730 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
3732 status
= nfserrno(error
);
3735 if (file_lock
.fl_type
!= F_UNLCK
) {
3736 status
= nfserr_denied
;
3737 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
3740 nfs4_unlock_state();
3745 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3746 struct nfsd4_locku
*locku
)
3748 struct nfs4_stateid
*stp
;
3749 struct file
*filp
= NULL
;
3750 struct file_lock file_lock
;
3754 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3755 (long long) locku
->lu_offset
,
3756 (long long) locku
->lu_length
);
3758 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
3759 return nfserr_inval
;
3763 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3767 &locku
->lu_stateowner
, &stp
, NULL
)))
3770 filp
= stp
->st_vfs_file
;
3772 locks_init_lock(&file_lock
);
3773 file_lock
.fl_type
= F_UNLCK
;
3774 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
3775 file_lock
.fl_pid
= current
->tgid
;
3776 file_lock
.fl_file
= filp
;
3777 file_lock
.fl_flags
= FL_POSIX
;
3778 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3779 file_lock
.fl_start
= locku
->lu_offset
;
3781 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
3782 nfs4_transform_lock_offset(&file_lock
);
3785 * Try to unlock the file in the VFS.
3787 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
3789 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3793 * OK, unlock succeeded; the only thing left to do is update the stateid.
3795 update_stateid(&stp
->st_stateid
);
3796 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3799 if (locku
->lu_stateowner
) {
3800 nfs4_get_stateowner(locku
->lu_stateowner
);
3801 cstate
->replay_owner
= locku
->lu_stateowner
;
3803 nfs4_unlock_state();
3807 status
= nfserrno(err
);
3813 * 1: locks held by lockowner
3814 * 0: no locks held by lockowner
3817 check_for_locks(struct file
*filp
, struct nfs4_stateowner
*lowner
)
3819 struct file_lock
**flpp
;
3820 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
3824 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
3825 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3836 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3837 struct nfsd4_compound_state
*cstate
,
3838 struct nfsd4_release_lockowner
*rlockowner
)
3840 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3841 struct nfs4_stateowner
*sop
;
3842 struct nfs4_stateid
*stp
;
3843 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3844 struct list_head matches
;
3848 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3849 clid
->cl_boot
, clid
->cl_id
);
3851 /* XXX check for lease expiration */
3853 status
= nfserr_stale_clientid
;
3854 if (STALE_CLIENTID(clid
))
3859 status
= nfserr_locks_held
;
3860 /* XXX: we're doing a linear search through all the lockowners.
3861 * Yipes! For now we'll just hope clients aren't really using
3862 * release_lockowner much, but eventually we have to fix these
3863 * data structures. */
3864 INIT_LIST_HEAD(&matches
);
3865 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3866 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3867 if (!same_owner_str(sop
, owner
, clid
))
3869 list_for_each_entry(stp
, &sop
->so_stateids
,
3871 if (check_for_locks(stp
->st_vfs_file
, sop
))
3873 /* Note: so_perclient unused for lockowners,
3874 * so it's OK to fool with here. */
3875 list_add(&sop
->so_perclient
, &matches
);
3879 /* Clients probably won't expect us to return with some (but not all)
3880 * of the lockowner state released; so don't release any until all
3881 * have been checked. */
3883 while (!list_empty(&matches
)) {
3884 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3886 /* unhash_stateowner deletes so_perclient only
3887 * for openowners. */
3888 list_del(&sop
->so_perclient
);
3889 release_lockowner(sop
);
3892 nfs4_unlock_state();
3896 static inline struct nfs4_client_reclaim
*
3899 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3903 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
3905 unsigned int strhashval
= clientstr_hashval(name
);
3906 struct nfs4_client
*clp
;
3908 clp
= find_confirmed_client_by_str(name
, strhashval
, use_exchange_id
);
3913 * failure => all reset bets are off, nfserr_no_grace...
3916 nfs4_client_to_reclaim(const char *name
)
3918 unsigned int strhashval
;
3919 struct nfs4_client_reclaim
*crp
= NULL
;
3921 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3922 crp
= alloc_reclaim();
3925 strhashval
= clientstr_hashval(name
);
3926 INIT_LIST_HEAD(&crp
->cr_strhash
);
3927 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3928 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3929 reclaim_str_hashtbl_size
++;
3934 nfs4_release_reclaim(void)
3936 struct nfs4_client_reclaim
*crp
= NULL
;
3939 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3940 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3941 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3942 struct nfs4_client_reclaim
, cr_strhash
);
3943 list_del(&crp
->cr_strhash
);
3945 reclaim_str_hashtbl_size
--;
3948 BUG_ON(reclaim_str_hashtbl_size
);
3952 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3953 static struct nfs4_client_reclaim
*
3954 nfs4_find_reclaim_client(clientid_t
*clid
)
3956 unsigned int strhashval
;
3957 struct nfs4_client
*clp
;
3958 struct nfs4_client_reclaim
*crp
= NULL
;
3961 /* find clientid in conf_id_hashtbl */
3962 clp
= find_confirmed_client(clid
);
3966 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3967 clp
->cl_name
.len
, clp
->cl_name
.data
,
3970 /* find clp->cl_name in reclaim_str_hashtbl */
3971 strhashval
= clientstr_hashval(clp
->cl_recdir
);
3972 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
3973 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
3981 * Called from OPEN. Look for clientid in reclaim list.
3984 nfs4_check_open_reclaim(clientid_t
*clid
)
3986 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
3989 /* initialization to perform at module load time: */
3992 nfs4_state_init(void)
3996 status
= nfsd4_init_slabs();
3999 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4000 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4001 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4002 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4003 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4004 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4006 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4007 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4008 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4009 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4011 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4012 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4013 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
4015 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
4016 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
4017 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
4019 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4020 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4021 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4023 memset(&onestateid
, ~0, sizeof(stateid_t
));
4024 INIT_LIST_HEAD(&close_lru
);
4025 INIT_LIST_HEAD(&client_lru
);
4026 INIT_LIST_HEAD(&del_recall_lru
);
4027 reclaim_str_hashtbl_size
= 0;
4032 nfsd4_load_reboot_recovery_data(void)
4037 nfsd4_init_recdir(user_recovery_dirname
);
4038 status
= nfsd4_recdir_load();
4039 nfs4_unlock_state();
4041 printk("NFSD: Failure reading reboot recovery data\n");
4045 get_nfs4_grace_period(void)
4047 return max(user_lease_time
, lease_time
) * HZ
;
4051 * Since the lifetime of a delegation isn't limited to that of an open, a
4052 * client may quite reasonably hang on to a delegation as long as it has
4053 * the inode cached. This becomes an obvious problem the first time a
4054 * client's inode cache approaches the size of the server's total memory.
4056 * For now we avoid this problem by imposing a hard limit on the number
4057 * of delegations, which varies according to the server's memory size.
4060 set_max_delegations(void)
4063 * Allow at most 4 delegations per megabyte of RAM. Quick
4064 * estimates suggest that in the worst case (where every delegation
4065 * is for a different inode), a delegation could take about 1.5K,
4066 * giving a worst case usage of about 6% of memory.
4068 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4071 /* initialization to perform when the nfsd service is started: */
4074 __nfs4_state_start(void)
4076 unsigned long grace_time
;
4078 boot_time
= get_seconds();
4079 grace_time
= get_nfs4_grace_period();
4080 lease_time
= user_lease_time
;
4081 locks_start_grace(&nfsd4_manager
);
4082 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4084 laundry_wq
= create_singlethread_workqueue("nfsd4");
4085 queue_delayed_work(laundry_wq
, &laundromat_work
, grace_time
);
4086 set_max_delegations();
4090 nfs4_state_start(void)
4094 nfsd4_load_reboot_recovery_data();
4095 __nfs4_state_start();
4101 nfs4_lease_time(void)
4107 __nfs4_state_shutdown(void)
4110 struct nfs4_client
*clp
= NULL
;
4111 struct nfs4_delegation
*dp
= NULL
;
4112 struct list_head
*pos
, *next
, reaplist
;
4114 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4115 while (!list_empty(&conf_id_hashtbl
[i
])) {
4116 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4119 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4120 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4124 INIT_LIST_HEAD(&reaplist
);
4125 spin_lock(&recall_lock
);
4126 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4127 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4128 list_move(&dp
->dl_recall_lru
, &reaplist
);
4130 spin_unlock(&recall_lock
);
4131 list_for_each_safe(pos
, next
, &reaplist
) {
4132 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4133 list_del_init(&dp
->dl_recall_lru
);
4134 unhash_delegation(dp
);
4137 nfsd4_shutdown_recdir();
4142 nfs4_state_shutdown(void)
4144 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
4145 destroy_workqueue(laundry_wq
);
4146 locks_end_grace(&nfsd4_manager
);
4148 nfs4_release_reclaim();
4149 __nfs4_state_shutdown();
4150 nfs4_unlock_state();
4154 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4155 * accessed when nfsd is starting.
4158 nfs4_set_recdir(char *recdir
)
4160 strcpy(user_recovery_dirname
, recdir
);
4164 * Change the NFSv4 recovery directory to recdir.
4167 nfs4_reset_recoverydir(char *recdir
)
4172 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
4176 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
4177 nfs4_set_recdir(recdir
);
4185 nfs4_recoverydir(void)
4187 return user_recovery_dirname
;
4191 * Called when leasetime is changed.
4193 * The only way the protocol gives us to handle on-the-fly lease changes is to
4194 * simulate a reboot. Instead of doing that, we just wait till the next time
4195 * we start to register any changes in lease time. If the administrator
4196 * really wants to change the lease time *now*, they can go ahead and bring
4197 * nfsd down and then back up again after changing the lease time.
4199 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4200 * when nfsd is starting
4203 nfs4_reset_lease(time_t leasetime
)
4205 user_lease_time
= leasetime
;